This proposal addresses statistical challenges that arise in HIV prevention research. Recent progress in prevention from different uses of antiretroviral drugs (in PrEP, microbicides, and treatment), male circumcision, and possibly vaccine has been noteworthy but modest. Combining across these modalities, however, might considerably enhance their effects. To address a crucial question facing policy makers--how best to deploy these modalities--requires knowledge of their combined effects and of the factors that impact on these effects. We propose to facilitate such research by developing efficient methods for analysis of data from observational and randomized studies of prevention interventions and for investigation of factors that may modify their degree of benefit. To do so, we propose to extend methods in variety of statistical areas, including causal inference and semiparametric theory, network analysis, and spatial analysis. We show how knowledge gained from new methods in each area can contribute to the usefulness of those in the other areas. Our first aim is to develop methods for using baseline covariates to improve efficiency of analyses of correlated outcome data that arise in group randomized trials (GRTs) and longitudinal studies; both types of studies are widely used in HIV prevention research. Because the number of experimental units in GRTs may be constrained by resource limitations, we develop methods appropriate for small as well as for large samples, and consider both individual- and group-level covariates. Estimation of network features that are important epidemic drivers from network samples is our second aim. This aim also considers methods for using such estimates in the construction of collections of networks--essential for realistic simulation of epidemics and of the impact of combination prevention interventions on them. These investigations will be useful for assessing whether intervention packages should be tailored to network features (and if so how), both through simulation and in analyses of data from research studies. The third aim develops methods for using data from HIV surveillance, especially of antenatal clinics, to estimate spatial and temporal variation in age-specific HIV prevalence and incidence. Such methods can aid in evaluating the impact of deployment of prevention interventions, especially those which are rolled out in sequentially in different regions. PUBLIC HEALTH RELEVANCE: This proposal addresses statistical challenges that arise in HIV prevention research. We focus on development of efficient methods for analysis of data from observational and randomized studies of prevention interventions and for investigating factors, like features of sexual networks, that impact on their success. We do so by extending methods in variety of statistical areas, including causal inference and semiparametric theory, network analysis, and spatial analysis. We show how knowledge gained from new methods in each area can contribute to the usefulness of the others in guiding deployment of prevention interventions.